Active-genetic neutralizing elements for halting or deleting gene-drives

Summary

CRISPR/Cas9-based gene-drive systems possess the inherent capacity to spread progressively throughout target populations. Here we describe two self-copying (or active) guide-RNA-only genetic elements, called e-CHACRs and ERACRs. These elements employ Cas9 produced in-trans by a gene-drive either to inactivate the Cas9 transgene (e-CHACRs) or to delete and replace the gene-drive (ERACRs). e-CHACRs can be inserted at various genomic locations and carry two or more gRNAs: the first copying the e-CHACR, and the second mutating and inactivating the Cas9 transgene. Alternatively, ERACRs are inserted at the same genomic location as a gene-drive, carry two gRNAs that cut on either side of the gene-drive to excise it. e-CHACRs efficiently inactivate Cas9, and can drive to completion in cage experiments. Similarly, ERACRs, particularly those carrying a recoded cDNA restoring endogenous gene activity, can drive reliably to fully replace a gene-drive. We compare the strengths of these two systems.

Graphical Abstract

eToc blurb

Xu et al. describes two genetic systems for neutralizing an active gene-drive that efficiently attenuate drive frequency in both pair crosses and cage population experiments. These neutralization systems either delete the gene-drive from the genome (ERACRs) or inactivate the Cas9 protein (e-CHACRs) to halt the gene-drive.

Introduction

Harnessing natural or synthetic gene-drives to bias inheritance of beneficial traits in populations was proposed over 60 years ago (Curtis, 1968), and variations on efficient “low-threshold” systems, such as homing endonucleases (Chevalier and Stoddard, 2001; Macreadie et al., 1985), have been modeled extensively over the past two decades (Burt, 2003; Deredec et al., 2008; Eckhoff et al., 2017). Recently, CRISPR/Cas9 genome editing tools have enabled development of several highly efficient gene-drive (or active genetic) systems in insects (Gantz and Bier, 2015; Gantz et al., 2015; Hammond et al., 2016; Kyrou et al., 2018; Li et al., 2019), yeast (DiCarlo et al., 2015), and bacteria (Valderrama et al., 2019). A mammalian gRNA-only “split-drive” prototype has also shown significant promise in the mouse (Grunwald et al., 2019).

CRISPR/Cas9-based gene-drives propagate by creating double-stranded DNA breaks at the precise site on the homologous chromosome where they are inserted into the genome. In the germline, the homology directed repair (HDR) pathway copies the gene-drive element into the break on the homologous chromosome, resulting in super-Mendelian transmission of that element to progeny. Mathematical modeling predicts that such efficient gene-drive systems should spread rapidly throughout a population following logistic growth dynamics, even when released at low seeding levels (Burt, 2003; Eckhoff et al., 2017; Gantz and Bier, 2016).

Discussion in the scientific literature, workshops, and the media have raised several potential concerns including scenarios in which a low-threshold system spreads beyond its intended zone of application (Adelman et al., 2017; James et al., 2018; Organisms, 2016; Warmbrod et al., 2020). One mitigation strategy for limiting gene-drive systems to a chosen region is to develop neutralizing genetic systems that eliminate or prevent further dissemination of the gene-drive.

We previously proposed two designs for self-copying (or “active”) neutralizing genetic elements that either inactivate Cas9 carried by a gene-drive (e-CHACR; [erasing] CHACR = Construct Hichhiking on the Autocatalytic Chain Reaction), or delete and replace the gene-drive (ERACR = Element Reversing the Autocatalytic Chain Reaction) (Gantz and Bier, 2016) (Fig. 1A). A key design feature of both elements is that they encode guide RNAs (gRNAs) but not Cas9. e-CHACRs can be inserted into the genome at any desired location and encode two or more gRNAs. One gRNA cuts at the genomic site of e-CHACR insertion, enabling self-copying in the presence of a transacting source of Cas9. The additional gRNA(s) target cleavage and inactivation of the Cas9 transgene component of a gene-drive element. ERACRs are inserted at the same genomic site as a gene-drive and encode two gRNAs that combine with Cas9 produced by the drive, to cut on either side of the drive element to delete and replace it.

Figure 1: Gene-drive and neutralizing drive elements.

A) Scheme depicting gene-drives and neutralizing elements. Left: MCR (mutagenic chain reaction) gene-drive element carrying Cas9, an eGFP fluorescent marker, and a gRNA- for copying. Center: eCHACR carrying two or more gRNAs: one gRNA (blue) for copying at its genomic insertion site, a second gRNA (purple) targeting Cas9, and a DsRed (or eGFP) fluorescent eye marker. e-CHACRs are typically inserted at a different chromosomal site (locus A) than the gene-drive (locus B). Right: ERACR carrying gRNAs that target sequences flanking the MCR-GFP element and a DsRed marker. B) Two MCR elements inserted at the same site in the y locus: 1) MCR lacking a fluorescent marker (third row), and 2) MCR-GFP, an eGFP-marked version (bottom row) carrying the same core components (vasa-Cas9 and gRNA-y1). C) Cross scheme for generating MCR-GFP F₁ “master females” and their F₂ progeny. Phenotypes of F₀, F₁, and F₂ progeny are depicted schematically. D) Percentage of GFP⁺ F₂ progeny (carrying MCR-GFP element) recovered per cross. Fly heads depict eye phenotypes determined by the white (w) locus: wild-type = w⁺ (red eyes), recessive w⁻ (white eyes), eye fluorescence markers (GFP = radiating green lines), and body color (y⁺ = brown, y⁻ = yellow).

When seeded into a gene-drive population, such gRNA-only driven neutralizing systems should follow the same logistic growth trajectory as a gene-drive released into a native population (Gantz and Bier, 2016). In the absence of a Cas9 source, however, these neutralizing elements are inherited in a standard Mendelian fashion, as when they either inactivate (e-CHACR) or eliminate (ERACR) a gene-drive element.

In this study, we test and analyze the activities of several e-CHACR and ERACR elements in Drosophila melanogaster (D. mel.). We find that while the e-CHACRs vary in their copying efficiency, all mutate and inactivate Cas9 efficiently, and the one tested drives to completion in population cages. Similarly, ERACRs often copy and delete a gene-drive element as intended, but can damage target chromosomes and generate various rare recombinant outcomes. Despite these imperfections, ERACRs, particularly those carrying functional recoded sequences that restore endogenous gene activity, can fully replace a gene-drive element in cage experiments. We discuss these results with regard to the potential utility of e-CHACRs to halt the spread of gene-drives.

Results

Active genetic elements such as gene-drives and gRNA-only neutralizing elements (Fig. 1A) bypass standard rules of inheritance imposed by independent chromosome assortment and linkage of nearby loci. Tracking such fluidly copying elements requires careful genetic bookkeeping by following each genetic element with a different fluorescent transgene and using genetic markers tightly linked to these elements to distinguish donor (chromosome of origin for a given active genetic element) versus receiver (target chromosome to which a drive element copies) chromosome homologs. For example, in this study, we used an eGFP-marked gene-drive (MCR-GFP) element inserted in the yellow (y) locus and DsRed-marked neutralization elements (e-CHACRs and ERACRs). We also marked donor versus receiver X chromosomes either with different alleles of white (e.g., white apricot = w^a; phenotype: orange eyes), located 1.5 centimorgans centromere proximal to y (Fig. 1B, 3A), or a viable allele of the achaete-scute locus (ac⁴; phenotype: missing innervated thoracic bristles) located immediately nearby (~9 kb centromere proximal to y: Fig. 2A).

Figure 3: ERACR construct designs and crossing schemes.

A) Diagrams of MCR-GFP and ERACR elements. Fly heads on the left indicate the phenotype of each strain. Schematic not drawn to scale. B) Cross schemes to generate F₁ “master females” with ERACRs in-trans to the MCR-GFP element and their F₂ progeny. Crossing schemes for y⁻ ERACR-min (left) and for y⁺ ERACR-1 and y⁺ ERACR-2 (right). Fly heads depict eye color (wild-type = w⁺ (red), w⁻ (white), or w^a (orange), eye fluorescence (radial emanating lines of eGFP (green), DsRed (red), both eGFP and DsRed (alternating green and red), or neither fluorescence (white), and body color: y⁺ (brown) or y⁻ (yellow). Expected F₂ phenotypes bolded and outlined with black boxes.

Figure 2: e-CHACR-wR versus the MCR-GFP element.

A) Schematic of DsRed⁺ e-CHACR-wR element linked to the ac⁴ allele (e-wR^ac) and the y⁺ ac⁺ marked MCR-GFP allele. B) Crossing scheme for testing the e-CHACR-wR (e-wR) against the MCR-GFP element. e-CHACR-wR females were mated to MCR-eGFP males to generate F₁ master females that were then pair-mated to ac⁴ (ac⁻) males. F₂ progeny were screened and analyzed for MCR-GFP presence and presence (DsRed⁺) or absence (w^NHEJ) of e-CHACR-wR on ac⁺ or ac⁴ marked chromosomes. C,C’) Percentage of fluorescence phenotypes in total female (C) or male (C’) F₂ progeny per cross. D,D’) Prevalence of GFP⁺ and GFP⁻ alleles in MCR-GFP donor (left) and receiver F₂ D) females or D’) males (right). E,E’) Prevalence of body color in total F₂ E) females or E’) males. F) Percentage of MCR-GFP donor (ac⁺, black dots) and receiver (ac⁻, pink dots) alleles in F₂ males.

Super-Mendelian transmission of the MCR-GFP gene-drive element

As a first step in our analysis we generated an eGFP-marked MCR-GFP gene-drive element carrying a Cas9 transgene expressed under control of the vasa promoter (vasaCas9) and a gRNA (gRNA-y1) directing its copying at the y locus (Gantz and Bier, 2015) using a genetic “tagging” method (Materials and Methods) (Fig. 1B). We assessed the drive performance of the MCR-GFP element in numerous parallel single-pair mating crosses in which so-called F₁ “master females” (females carrying both gRNA and Cas9 transgenes) were crossed to y⁺ w⁻ males (Fig. 1C). A substantial fraction (~20–35%) of crosses resulted in 100% transmission of the MCR-GFP element to all F₂ progeny (Fig. 1D). Overall, the MCR-GFP element exhibited an average super-Mendelian transmission rate of ~85%. Similar results were obtained in crosses generated from master females that inherited the drive paternally (Fig. 1D) or maternally (Figs. S1A,B), albeit with somewhat reduced efficiency in maternal crosses. A noteworthy feature of these experiments, revealed most obviously in histogram summaries of many individual pair matings (Fig. S1C–D”), was that the frequencies of F₂ progeny inheriting the gene-drive element were distributed in a non-Gaussian fashion. Potential explanations for this Poisson-like distribution of gene-drive transmission are considered in Data S1. Also, nearly all F₂ progeny displayed a y⁻ mutant phenotype whether not they carried the MCR-GFP drive, consistent with both zygotically and maternally provided Cas9/gRNA complexes acting on and mutating the wild-type paternal y⁺ allele, a phenomenon that we and others have previously documented (Champer et al., 2017; Gantz and Bier, 2015; Gantz et al., 2015; Guichard et al., 2019; Hammond et al., 2016; Li et al., 2019; Lin and Potter, 2016).

Experiments with w^a marked receiver chromosomes (Fig. S2A) revealed that the two chromosome homologs were inherited following standard Mendelian segregation (Fig. S2B). We conclude that the great majority of the super-Mendelian inheritance displayed by the MCR-GFP element in such crosses can be attributed to gene-drive copying events rather than to biased transmission of donor versus target receiver chromosomes.

e-CHACRs efficiently inactivate Cas9, but copy with a range of frequencies

We inserted e-CHACRs into loci on different chromosomes: the X-linked white locus (e-CHACR-wG: eGFP marked; e-CHACR-wR: DsRed marked), and the third chromosome ebony (e-CHACR-e) and knirps (e-CHACR-k) loci (see online Data S1 file for further details on e-CHACR analysis). Strains carrying these e-CHACRs were crossed to those harboring a standard “static” (non-copying) vasaCas9 source (Figs. S3–5,S10) or the active MCR-GFP gene-drive inserted at y (Fig. 2; Figs. S7–9,S11) to generate F₁-generation master females bearing both elements. The phenotypes of F₂ progeny from many parallel pair-mating crosses between F₁ master females and males of informative genotypes (i.e., y^±, w^±, ac^±) were tabulated, as discussed below. We evaluated two key e-CHACR performance parameters: 1) the efficiency of anti-Cas9 gRNAs (gRNA-C1, carried by the w and kni e-CHACRs; or gRNA-C3, carried by the ebony e-CHACR) in mutating and inactivating Cas9, and 2) the rate of e-CHACR copying to the target receiver chromosome.

Cas9 inactivation by e-CHACRs was assessed in several ways. For e-CHACRs inserted at w (Fig. 2; Figs. S3,S4,S7), we monitored a highly penetrant Cas9-dependent somatic mosaic eye phenotype (eyes consist of red and white sectors). By this mosaic eye metric, we estimate that Cas9 was inactivated in ~99% of F₂ progeny by either gRNA-C1 or gRNA-C3, both of which target sites in the Cas9 transgene encoding catalytically essential amino acids. Non-mosaic F₂ progeny carrying e-CHACR-wG and the presumably mutated Cas9 source transmitted the GFP⁺ element to F₃ progeny at standard Mendelian ratios (Fig. S3E), in contrast to the high Super-Mendelian frequencies observed in F₂ progeny (Figs. S3C,C’). In experiments involving the MCR-GFP drive (Fig. 2; Figs. S7;S9;S11) we also assessed the full-body y⁻ pigmentation phenotype, which as mentioned above, is fully penetrant in F₂ progeny of master females carrying an active MCR-GFP element. 100% of F₂ progeny had y⁺ phenotypes, revealing that Cas9 activity was eliminated by all e-CHACRs. We also sequenced genomic DNA isolated from individual F₂ progeny, confirming that the gRNA target sites had been mutated in all samples analyzed from crosses with e-CHACR-wg (carrying gRNA-C1) and e-CHACR-e (carrying gRNA-C3) (Fig. S6).

We observed a range of e-CHACR transmission frequencies. Copying was highly efficient (95–99% transmission to F₂ progeny) for both GFP and DsRed-marked e-CHACRs inserted at w (e-CHACR-wG and e-CHACR-wR, respectively)(Fig. 2; Figs. S3–5). e-CHACR-k, inserted in the knirps (kni) locus, exhibited intermediate levels of copying (87–90% transmission; Fig. S11), while e-CHACR-e, inserted at ebony (e), was transmitted at Mendelian frequencies (Fig. S9). The absence of drive for e-CHACR-e was puzzling since previous experiments with other split-drive elements indicated that the same gRNA-e1 carried by e-CHACR-e sustained modest super-Mendelian copying (López Del Amo et al., 2019). We speculated that efficient mutagenesis of the Cas9 transgene by gRNA-C3 might prevent e-CHACR-e from copying. Indeed a cleavage-resistant form of Cas9 (Cas9*) that could not be targeted by gRNA-C3 sustained modest copying of e-CHACR-e (~60% transmission) (Fig. S10), similar to that previously reported for the split system driven by the same gRNA-e1 (López Del Amo et al., 2019). We conclude that all three e-CHACRs are highly efficient at eliminating Cas9 activity with either of two Cas9 targeting gRNAs, and that their differing ability to copy most likely reflects the relative activities of the gRNAs that sustain their copying (see Data S1 and Discussion for further analysis).

e-CHACRs efficiently inactivate the MCR-GFP drive

We tested the ability of the three different e-CHACRs to inactivate the MCR-GFP full-drive element, and all functioned efficiently to mutate Cas9 as judged by the virtual absence of F₂ individuals displaying mosaic eye (>99%) or full-body y⁻ (100%) phenotypes (Fig. 2E,E’; Fig. S9E,E’; Fig. S11E,E’). All three e-CHACRs also markedly reduced the frequency of MCR-GFP copying, albeit to varying extents. e-CHACR-wR reduced MCR-GFP copying by 5–10 fold (i.e., from ~70%: Fig. 1D to 7–13%: Fig. 2D,D’; Fig. S7D,D’), e-CHACR-e by ~ 5-fold (Fig. S9D,D’), and e-CHACR-k by ~2-fold (Fig. S11D,D’). We conclude that e-CHACRs efficiently inactivate Cas9, copy themselves in the presence of a full-drive element or static sources of Cas9, and significantly reduce copying of the MCR-GFP gene-drive element (see Discussion). These various reinforcing activities likely contribute to the efficient performance of e-CHACR-wR in population cage experiments described below.

e-CHACRS can preferentially bias inheritance of donor chromosomes

Visible markers (e.g., w^a, ac⁴) closely linked to the e-CHACR and MCR-GFP elements, permitted us to determine whether the donor and receiver chromosome homologs segregated in a biased fashion. We observed two types of significant bias in chromosomal inheritance. First, when the X-linked e-CHACR-wG targets a static Cas9 source inserted at the closely linked y locus, we observed a nearly 2:1 bias in the inheritance of the donor e-CHACR chromosome over the receiver chromosome in male (Fig. S3C’) but not female (Fig. S3C) F₂ progeny, which was accompanied by a corresponding excess of females to males (Fig. S8B,D,F). In contrast, when autosomal e-CHACR-e targeted a nearby Cas9-GFP source at w, the single-cut target chromosome was inherited at Mendelian frequencies in F₂ progeny of both sexes (Fig. S10A–C). Also, e-CHACR-wR driven by an autosomal source of Cas9 did not bias the sex of recovered F₂ progeny (Fig. S5). The handicap against receiver chromosomes in the former crosses, however, did not permanently damage target chromosomes since F₂ females carrying mutated Cas9 chromosomes transmitted them to ~50% of their F₃ progeny (Fig. S3E).

The second type of biased chromosome inheritance occurred when any of the three e-CHACRs targeted the MCR-GFP element. In these scenarios, the same locus on both chromosome homologs is targeted for cleavage (the donor MCR-GFP allele with anti-Cas9 gRNAs, and the receiver target allele with gRNA-y1 expressed by the MCR-GFP element). We observed excess transmission of y⁺ac⁴ receiver chromosomes in F₂ males, but not females (Fig. 2F; Figs. S7F;S9F;S11F). We also recovered an unexpected class of GFP⁻ donor chromosomes (Fig. 2D,D’; Figs. S7D,D’;S9D,D’). Of 17 such isogenized GFP⁻ target chromosomes tested, 10 (59%) carried male-lethal alleles, and all could be rescued by a duplication covering the y locus (Fig. S12A; Table S1). As analyzed in greater detail in Data S1, we conclude that at least two different types of chromosome bias can be induced by e-CHACR elements: one, seen when cutting chromosomes carrying a static Cas9 source twice, does not irreparably damage the receiver chromosome. The second, associated with e-CHACRs targeting the MCR-GFP drive element, can induce local damage to the donor chromosome resulting in male lethality (or homozygous lethality in females).

ERACR versus MCR-GFP crossing schemes

We tested three DsRed-marked ERACR constructs designed to delete and replace the MCR-GFP element inserted at y (see online Data S2 for in-depth analysis of ERACRs). All three ERACRs carry two gRNAs (gRNA-y2 and gRNA-y3) that direct Cas9 cleavage to either side of the MCR-GFP element (Fig. 3A). ERACR-min carries just the aforementioned minimal elements (DsRed and two gRNAs) while ERACR-1 and ERACR-2 also carry recoded y cDNA coding sequences at their 5’ junction that seamlessly restore function of the y locus. ERACR-2 shares fewer homologous sequences with the MCR-GFP element than ERACR-1. y cDNA sequences are fully recoded for ERACR-2, whereas only the 5’ junction is recoded in ERACR-1. Also, U6-promoter sequences driving expression of gRNAs carried by ERACR-2 derive from a distantly related Drosophilid (D. grimshawi), which have little sequence homology to those carried by the MCR-GFP element, and are oriented in the opposite direction to those carried on the gene-drive to minimize potential spurious recombination events between the two elements.

Fly strains carrying the different DsRed⁺ ERACR constructs were crossed to a w^a marked MCR-GFP strain (Fig. 3B; see also Fig. S13 and Data S2 for additional crossing schemes and analysis). F₁ master females carrying the w^a MCR-GFP chromosome and w⁻ ERACR elements were crossed to w⁻ males and their F₂ progeny scored for fluorescence, eye color, and whole-body pigmentation phenotypes. This scheme permits w⁻ ERACR-bearing donor chromosomes to be distinguished from w^a MCR-GFP target chromosomes.

ERACRs frequently delete and often replace a gene-drive element

Three prominent outcomes were observed among progeny inheriting the w^a receiver chromosome: 1) deletion and replacement of the MCR-GFP element with the ERACR (phenotype: DsRed⁺,GFP⁻, Fig. 4A–C); 2) retention of the MCR-GFP element (phenotype: DsRed⁻,GFP⁺, Fig. 4D,E) with NHEJ induced indels at both the gRNA-y2 and gRNA-y3 cut sites (i.e., both ERACR gRNAs direct efficient target cleavage); and 3) deletion of the MCR-GFP element without copying the ERACR (phenotype: DsRed⁻,GFP⁻, Fig. 5A–E), a category only abundantly observed in females. Comprehensive analysis of the various ERACR versus MCR-GFP outcomes is presented in Fig. 4 and Fig. 5 and Figs. S14–17A and in-depth analysis of these events is presented in Data S2).

Figure 4: ERACRs delete and replace the MCR-GFP drive.

Phenotypic frequencies and deduced gene conversion events in F₂ progeny from crosses depicted in Fig. 3A. (additional crossing schemes in Figs. S13A–E). Black type = mean of percentages across all vials and orange type = estimated receiver conversion frequencies (e.g., DsRed w^a males/total w^a males) (see Data S3 for details). A) DsRed⁺ inheritance is a proxy for scoring ERACR prevalence (DsRed⁺, GFP⁻ and DsRed⁺, GFP⁺ progeny included). A’) The subset of data plotted in panel 4A with traceable donor and receiver chromosomes re-plotted by donor (w⁻; left graph) versus receiver (w^a; right graph) chromosomes. B) Proportion of F₂ males or females inheriting the donor (w⁻) versus receiver (w^a) chromosome. C,E) Schematics illustrating predicted gene conversion events responsible for specific phenotypes: sequences of relevant junctions shown in (Fig. S14). D) GFP inheritance is a proxy for MCR-GFP prevalence (DsRed⁺,GFP⁺ plus DsRed⁻,GFP⁺ F₂ progeny). E) MCR-GFP alleles, although intact, have NHEJ-induced indels at the y2 and y3 cut sites.

Figure 5: Alternative ERACR versus MCR-GFP outcomes.

Partial copying or fusion of sequences on the receiver chromosome. A) Non-fluorescent DsRed⁻,GFP⁻ F₂ progeny as a proxy for MCR-GFP deletion events. A’) Data in panel A re-plotted by inheritance of y⁻ (lethal deletion) versus y⁺ (recombination) alleles. Body color permits inference of the type of excision event that occurred, as depicted in C-E. B) Loss of essential sequences distal to the ERACR gRNA cut sites result in male-lethal alleles. Viability of several such alleles can be rescued in males by duplications covering the tip of the X chromosome (Fig. S12B). C-E) The MCR-GFP element is deleted, but not replaced by ERACR, producing three distinct observed outcomes. C) Hypothesized repair mediated by partial pairing of un-recoded y sequences carried by ERACR-1 and endogenous y sequences 3’ to the MCR-GFP element result in expression of the recoded y cassette and a wild-type body color. D) NHEJ events joining adjacent sequences at the gRNA-y2 and gRNA-y3 cut sites. E) Likely pairing between 17 bp of the gRNA-y2 genomic target sequences 5’ to the MCR-GFP with correspondingly oriented sequences in the gRNA-y2 transgene carried by either ERACR-min or ERACR-1. F) Prevalence of DsRed⁺, GFP⁺ F₂ progeny in which MCR-GFP and ERACR sequences are both present on the receiver chromosome. G) Two examples of MCR-GFP/ERACR fusion events (see Fig. S14).

Several salient features emerged from these experiments. First, the three ERACRs performed similarly overall (Figs. 4,5; Fig. S15), deleting and replacing the MCR-GFP element (~31%), leaving the MCR-GFP in place (~24%), and, in females, deleting the MCR-GFP without copying (~43%). There were also various rare outcomes (~1%) based on illegitimate recombination events, which differed based on ERACR design (discussed further in Data S2). Second, the DsRed⁻, GFP⁻ phenotype was recovered >10-fold more often in females than males (Fig. 5A,A’). The great majority of such alleles were male lethal, resulting in a 2:1 excess of female over male F₂ progeny (Fig. S17A). In all tested cases, lethality of these DsRed⁻,GFP⁻ alleles could be rescued by a duplication of the tip of the X chromosome covering the y locus (Fig. S12B). These findings indicate that ERACRs often damaged the target chromosome, disrupting essential functions in the vicinity of the target locus. Finally, although illegitimate recombinant events were rare, they were recovered (Fig. 5C,E,F,G). A few events lead to the generation of DsRed⁺,GFP⁺ mosaic elements (Fig. 5F,G) that retained some capacity to copy all (Fig. S20A–C’) or portions (Fig. S20B–D’) of the elements. Also noteworthy, the frequency of such outlier events was significantly reduced for ERACR-2 relative to ERACR-1 (Fig. 5A,A’,F; Figs. S14,S15), indicating that elimination of homologous sequences between ERACR and gene-drive targets is a favorable design strategy.

One-sided homology mismatch underlies ERACR-induced chromosome damage

Like e-CHACRs targeting receiver chromosomes for dual cleavage, ERACRs favor inheritance of donor chromosomes, suggesting that ERACR-induced chromosome damage might result from cutting the receiver chromosome twice at neighboring sites. We tested this hypothesis by constructing single-cut versions of ERACR-2, which should not generate damage according to the double-cut model. Single-cut ERACRs carried either only gRNA-y2 (ERACR2-y2: cutting 5’ and centromere distal to the MCR-GFP element), or only gRNA-y3 (ERACR2-y3, cutting 3’ and centromere proximal to the MCR-GFP element) (Fig. 6A). We tested these elements for super-Mendelian transmission as well as induction of damage to the target chromosome.

Figure 6: Drive performance of single-cut ERACRs versus MCR-GFP drive.

A) Schemes illustrating single-cut and double-cut ERACR designs. B) Crossing scheme for generating and testing transmission by single-cut w^a ERACR/MCR-GFP F₁ master females. C-E) Single-cut versions of ERACR-2 are placed in-trans to the MCR-GFP element where the gRNA-y2 and gRNA-y3 are separated by a distance of 11.3 kb (see also Fig. S19 for Copy-Cat analysis). C) Fluorescent phenotypes for single-cut DsRed⁺ ERACR2-y2 and ERACR2-y3 and MCR-GFP. D) Percent DsRed⁺ females or males inheriting either the donor (w⁻) ERACR-2 single-cut chromosome or the receiver (w^a) chromosome. E) Donor versus receiver chromosome transmission for single-cut ERACR2-y2 and ERACR2-y3.

Copying of single-cut ERACRs was assessed by pair-mating them to w^a MCR-GFP individuals according to the scheme depicted in Fig. 6B, which is analogous to that used for the double-cut ERACRs (Fig. 3B). Analysis of fluorescence associated with w^a receiver chromosomes in F₂ progeny (Fig. 6C,D) revealed that the frequency of deleting and replacing the MCR-GFP element was significantly reduced for both ERACR2-y2 (avg. = 4.9%, Fig. 6D) and ERACR2-y3 (avg. = 1.7%, Fig. 6D) relative to the dual-cutting ERACR-2 (avg. = 16%, Fig. 4A’, right panels), even when combined.

Crosses of ERACR2-y3 to the MCR-GFP element also reveal whether these single-cut elements might damage the target chromosome. We observed a prominent category (17%) of DsRed⁻ GFP⁻ F₂ female progeny that was absent in male siblings (Fig. 6C), which is indicative of deleting the MCR-GFP element without copying the ERACR. This rate of DsRed⁻,GFP⁻ progeny is similar to that observed for the double-cut ERACR-2 (22.7% in F₂ females, versus 1.1% in males: Fig. 5A), suggesting that single-cut ERACR2-y3 damages the target chromosome at a frequency comparable to that of the double-cut elements. Although crosses of ERACR2-y2 to the MCR-GFP element did not produce a prominent class of DsRed⁻,GFP⁻ F₂ female progeny, we did observe a similar ~10% disparity in the fraction of GFP⁺ female versus male progeny. This finding could be explained by chromosome damage being induced distal to the gRNA-y2 cut site, leaving the centromere-proximal GFP transgene intact. Both single-cut-ERACRs also lead to an excess of females over males F₂ progeny inheriting the receiver chromosome (Fig. S17B) as did all double-cut ERACRs (Fig. S17A). These results suggest that both single-cut ERACRs damage the target chromosome at appreciable frequencies and that damage is localized centromere-distal to the gRNA-directed cleavage sites.

We also compared the frequencies with which single- versus double-cut ERACRs copied and induced chromosome damage when confronting a wild-type y⁺ allele in a so-called “copy-cat” configuration with a static source of Cas9 provided in-trans (Figs. S18;S19). Here, the single-cut ERACR2-y2 (Fig. S19B) copied nearly as well as the full double-cut ERACR-2 (Fig. S19A) as judged by the proportion of DsRed⁺,GFP⁺ progeny (~20%). ERACR2-y3 copied less well, albeit notably better (~6%) than when challenged with the MCR-GFP (~1.5%). Both single- and double-cut ERACRs again induced comparable reductions in transmission of the GFP-marked target chromosome in F₂ males, but not females, consistent with damage to receiver chromosomes. We conclude that the likely basis for the majority of ERACR-induced chromosomal damage is lack of DNA homology on one side of the ERACR (the side missing the second gRNA), rather than cutting the target chromosome twice. We also note that single-cut ERACRs copy better with less distance between flanking homology sequences (e.g., 2.2 kb in a copy-cat versus 11.3 kb MCR-GFP mode).

ERACR and e-CHACRs do not sustain shadow drive

We previously documented a maternal “shadow-drive” phenomenon in individuals descended from Cas9-bearing mothers that inherit a gRNA-based drive element but not the Cas9 transgene (Guichard et al., 2019). Since both e-CHACRs and ERACRs eliminate Cas9 activity, we wondered whether this early action might preclude accumulation of sufficient Cas9/gRNA stores to sustain shadow-drive in the subsequent generation. We examined the drive potential of e-CHACR or ERACR elements in F₂ progeny of F₁ master females lacking the Cas9 transgene and thus carrying only maternal stores of the endonuclease provided by their F₁ mothers (Fig. S21A,C). We tested e-CHACR or ERACR transmission to F₃ progeny from such F₂ females and observed Mendelian frequencies of GFP transmission (Fig S21,B,D–F), revealing that neither neutralization element sustained significant shadow-drive.

e-CHACR-wR inactivates the MCR-GFP gene-drive in population cages

Results from pair-mating crosses revealed that e-CHACR-wR efficiently eliminated Cas9 activity and copied efficiently when combined with either static (Figs. S3;S4;S5) or gene-drive (Figs. 2; Fig. S7) borne sources of Cas9. We generated mathematical models based on these single-generation data (Data S3) and fitted parameters based on observed time-series data from the cage trials (Fig. S22, Fig. 7A - solid lines). These initial values were then used to run simulations, shown as arrays of pale-colored lines, matching the schemes for the different displayed phenotypes (Fig. 7B). Simulations with the set of fitted parameters (Data S3, Table Smod4) were largely consistent with experimental values for the frequencies of different phenotypic classes observed in the pair-mating crosses. We were also able to infer parameters revealing informative deviations from assumptions of random mating between all genotypes. In particular, the modeling indicated a relatively high degree of assortative mating within groups having shared eye pigmentation (i.e., w⁺ females preferentially mating with w⁺ males over w⁻ males).

Figure 7: Cage experiments: MCR-GFP versus e-CHACR-wR and ERACRs.

A,C,E) Modeling of MCR-GFP versus e-CHACR-wR (A) and ERACR (C,D) dynamics (solid lines: same as B,D,E) and model fits (dotted lines) plotted for the frequencies of MCR-GFP (green) and e-CHACR-wR or ERACRs (red). B) Plot of fraction of individuals with different phenotypes over 12 generations. Green = MCR-GFP and Red = e-CHACR-wR prevalence in the total population (e.g., both males and females). Orange = females carrying both elements. Yellow = females with mosaic eyes indicative of Cas9 activity. Dark traces represent separate cage replicates and pale lines = model simulations (also in panels D,F). C,E) Modeling of MCR-GFP versus ERACR dynamics over 26 generations (C) and ERACR-2 (E). D,F) ERACR-min (D) and ERACR-2 (F) versus MCR-GFP cage experiments. Red = DsRed⁺ ERACRs. Green = MCR-GFP⁺. Yellow = both markers.

In conjunction with the modeling above, we competed e-CHACR-wR against the MCR-GFP element in population cages. We seeded triplicate cages with equal numbers of homozygous e-CHACR-wR and MCR-GFP individuals. At each generation (n) we scored half of the individuals per cage (approximately 150 random flies) for their fluorescent and y^± phenotypes, while the other half was used to seed the next generation (n+1) of cages. In all three cages, e-CHACR (red curves) drove to completion in 9 generations (Fig. 7A,B). In addition, we observed a transient appearance and then disappearance (after 8 generations) of a population of DsRed⁺, GFP⁺ females that exhibited mosaic eye phenotypes (yellow curves in Fig. 7B), which reveals Cas9 activity when the two elements first encounter each other (e.g., as in master females). Outcrosses of sampled DsRed⁺, GFP⁺ males and females from generations 9 and 10 revealed that all tested MCR-GFP alleles had lost Cas9 activity (judged by the absence of female progeny with mosaic eyes - Fig. S22B,B’). In addition, 100% of progeny from crosses of DsRed⁺,GFP⁺ females from generations 9 and 10 to wild-type males were DsRed⁺,GFP⁻ indicated that the e-CHACR-wR had achieved 100% homozygous introgression (Fig. S22C,C’). Overall, simulations derived from these assumptions matched well with the observed efficient experimental performance of the e-CHACR (Fig. 7B).

Another multi-generational trend was that following an initial expected increase in the MCR-GFP frequency, which peaked at generation 7, prevalence of this element declined precipitously (green curve in Fig. 7B). Although a similar trend of initial surge followed by waning was observed in control crosses of the MCR-GFP element to y⁻ mutants (which harbor an intact gRNA-y1 cleavage site) (Fig. S22A), the peak was earlier (generation 5), and the decline more gradual. These later observations suggest that the MCR-GFP element may carry a fitness cost (hence its slow loss in control experiments). The e-CHACR may accelerate clearance of the MCR-GFP element, potentially by inaccurate repair of the targeted Cas9 transgene.

ERACRs replace the MCR-GFP gene-drive in population cages

Mathematical modeling based on single generation pair-mating ERACR data (Fig. S15) suggested that these neutralizing elements should behave as designed to eliminate the gene-drive element over multiple generations (Fig. 7C,E; Fig. S23F,F’,G,G’). In these simulations we made the following two simplified assumptions: 1) MCR-GFP elements retained after confrontation with an ERACR are immune to further action by ERACRs (due to generation of cleavage-resistant NHEJ mutations at both cut sites); and 2) double-negative (DsRed⁻,GFP⁻) alleles generated by ERACRs deleting, but not replacing, the MCR-GFP element, are fully viable in heterozygous females, but lethal in homozygous or hemizygous conditions. In addition, consistent with prior classic studies (Barker, 1962; Bastock, 1956; Diederich, 1941; Merrell, 1949; Sturtevant, 1915), we modeled a range of potential fitness costs for y⁻ versus ERACR-2 (y⁺) males based on y⁺ females preferring to mate with y⁺ males, as well as an apparent fitness cost associated with the MCR-GFP gene-drive. We confirmed a potent assortative mating preference of y⁺ females in single generation control crosses with our own stocks (Fig. S23A–D), and in competitive multi-generational cage experiments (Fig. S23E). Sampling over a broad range of parameters suggested that ERACR-min (y⁻) and, particularly ERACR-2 (y⁺), should produce sufficient drive in mixed populations carrying the MCR-GFP and ERACR (but no wild-type alleles) to replace and ultimately eliminate the gene-drive over several generations (Fig. 7C,E).

We conducted 3–4 separate cage replicates per ERACR composed of 25% homozygous ERACR and 75% homozygous MCR-GFP virgin male and female individuals in each cage. The frequency of the DsRed⁺ ERACR cassette for the ERACR-min experiment tripled in ~4 generations and then gradually approached fixation over the next 5 generations, with a modest degree of variation (~10–15%) exhibited between the four cages tested (Fig. 7D). Reciprocally, the frequency of the MCR-GFP marker decreased until it was nearly eliminated by generation 9. Consistent with random mating among y⁻ ERACR-min and MCR-GFP individuals, 30–35% of progeny from the first generation were trans-heterozygotes carrying both fluorescent markers. This double-positive (GFP⁺,DsRed⁺) population peaked in generation 3 and then steadily diminished. In contrast, the proportion of DsRed individuals in the ERACR-2 cages (Fig. 7F) remained approximately constant for one generation and then increased steeply and in tight synchrony, reaching fixation in 5–6 generations, with concomitant reduction and then complete elimination of the MCR-GFP element. Notably, ERACR-2 replicats displayed little variation in their drive trajectories. These experimental results match well with simulations (pale lines in Figs. 7D,F) based on parameters identified by model fitting (dotted lines in Figs. 7C, E), capturing key features of the overall kinetics and inter-trial variations.

Although there was an excellent overall fit between the observed data and modeling (Fig. 7C,E; Fig. S23F–G’), two ERACR-min replicates (#1 and #4) pursued moderately different trajectories from the other two (#2,#3), with the latter two matching the predicted modeling (Fig. 7C,E). Since no such events were observed in any of the y⁺ ERACR-2 replicates, these outlier trajectories may result from formation of y⁻ resistance alleles that would have been selected against in the ERACR-2 replicates due to assortative mating, but that could persist and potentially spread in the y⁻ ERACR-min replicates where they were on even footing. In addition, modeling and cage studies (Fig. S22) suggest that a fitness cost is associated with the MCR-GFP element, which might be aggravated by cryptic ERACR-induced damage by gRNA-y2.

We also competed the y⁺ ERACR-2 against a y⁻ NHEJ-induced point allele generated at the gRNA-y1 cleavage site (i.e., the same gRNA driving the MCR-GFP element) in the absence of Cas9 to isolate the effect of assortative mating (Fig. S23E). Over ~10 generations, chromosomes carrying the y⁺ ERACR-2 successfully overtook the y⁻ NHEJ population with typical inter-cage variation (Fig. S23E). These observations confirm that the y⁺ genotype has an appreciable competitive advantage over y⁻ in a cage setting. Also consistent with y⁺ ERACR-2 females mating preferentially with y⁺ ERACR-2 males, the frequency of DsRed⁺,GFP⁺ (ERACR-2 + MCR-GFP) flies in first generation progeny (Fig. 7F, yellow curves) was considerably lower (15–20%) than that predicted by random mating (37.5%) or observed for ERACR-min (Fig. 7D). We conclude that the modeling predictions conform closely to the experimental outcomes and that ERACRs, particularly those carrying a recoded transgene restoring function of a functionally important gene, have the potential to eliminate and replace a gene-drive element even once it has attained fixation in a population.

Discussion

Overall, both e-CHACRs (Fig. 8A) and (ERACRs) (Fig. 8B). offer promise for fulfilling their purpose of countering a gene-drive system (Fig. 8C–D), however, it is also important to take into account some of their unintended actions.

Figure 8: Summary Diagrams.

A) Diagram illustrating the generic action of the e-CHACRs against the MCR-GFP gene-drive. B) Diagram outlining the structures and homologous sequences serving as sites for potential SDSA-mediated partial copying between ERACR-2 and the MCR-GFP constructs (lightly shaded parallelograms). y1*, y2* and y3* indicate the loss of the gRNA-y1, gRNA-y2 and gRNA-y3 cut sites accompanying genomic insertion of the MCR-GFP element (gRNA-y1) or ERACRs (gRNA-y2 and gRNA-y3). C) Pie charts summarizing key e-CHACR performance parameters. D) Pie chart summary of F₂ male and female progeny outcomes derived from F₁ ERACR-2/MCR-GFP master females separated by inferred donor (w⁻ = red shaded sectors) versus receiver (w^a: peach = ERACR copied, green = MCR-GFP retention; gray =MCR-GFP deleted) chromosomes.

Drive performance of drive and neutralization systems

Several factors contribute to levels of drive we observed, which ranged in transmission frequencies from >95% (e-CHACR-wG/R inserted into the w locus) to only ~50% Mendelian inheritance (e-CHACR-e, but see below). e-CHACR-k and the MCR-GFP drive element copied at intermediate rates (~85%), which for the MCR-GFP is somewhat lower than estimated in prior experiments where transmission of an unmarked MCR element was estimated based on full yellow body phenotypes (Gantz and Bier, 2015). Since a significant fraction of crosses using the marked MCR-GFP element transmitted it to 100% of progeny, this may also have been the case in those original experiments, which were based on only a small number of crosses. Alternatively, it is possible that we overestimated gene conversion (e.g., the mosaic group might not have inherited the MCR). The size of the drive cassette could also impact copying efficiency since we have noted higher inheritance rates for a smaller drive element using the same gRNA-y1 (e.g., (López Del Amo et al., 2019), which was in the 90% range.

In pair-mating crosses, all three ERACRs deleted the target MCR-GFP gene-drive element from 74% of target chromosomes and copied themselves in its place as intended in 31% of those cases. The combined effect of these two outcomes is pronounced super-Mendelian inheritance of ERACR alleles in F₂ males (85–90%). In females, only copying events contributed to biased inheritance of the ERACR (e.g., ~65%), since male-lethal MCR-deleted chromosomes can survive in the heterozygous state (~20% of total females). In multi-generational contexts, however, damaged alleles carried by females are rapidly purged from the population (see below).

As noted above, e-CHACRs copied at different rates, which most likely reflects relative gRNA efficiencies. Two cases are informative in this regard. First, e-CHACR-e did not copy when combined with Cas9, although in a split-drive configuration, without the Cas9-targeting gRNA, it exhibited modest drive (~60%) (López Del Amo et al., 2019). When combined with a cleavage resistant Cas9* form, however, e-CHACR-e drove at levels similar to those in the split-drive experiments. We hypothesize that a “weak” gRNA, such as that carried by e-CHACR-e, can fail entirely if the source of Cas9 is eliminated prematurely. Similarly, the intermediate level of MCR-GFP drive (~85%) was reduced 2–10 fold when combined with various e-CHACRs, again suggesting that early elimination of Cas9 activity can reduce the performance of suboptimal gRNAs. In contrast, when the e-CHACR-wG/Rs carrying an efficient gRNA were combined with static or MCR-GFP sources of Cas9, they copied at rates equal to those observed in split-drive experiments (López Del Amo et al., 2019). Thus, choosing an efficient gRNA seems to be an important design feature for e-CHACR elements.

ERACRs carry two gRNAs targeting the receiver chromosome at two neighboring sites. The two gRNAs increase the rate of ERACR copying relative to single-cut ERACRs that carry only one or the other gRNA - suggesting some form of synergy even though each gRNA was expected to act independently. This disparity was significantly more pronounced when the two gRNA cut sites were spaced further apart (e.g., when the MCR-GFP element is inserted between them).

Neutralizing elements can bias chromosomal inheritance

Marking donor versus receiver chromosomes permitted us to discern three different types of biased inheritance of these chromosome homologues. First, e-CHACRs cutting the same chromosome twice generated a 2:1 bias in favor of transmitting the donor X chromosome in males, but not females. This bias did not lead to any obvious permanent damage, however, since females segregated both donor and receiver chromosomes to their progeny with equal frequency (Fig. S3E). Second, e-CHACRs generated a class of donor MCR-GFP alleles lacking the internal GFP marker, over half of which were male lethal. In this scenario, gRNAs from e-CHACRs (anti-Cas9 gRNAs) and MCR-GFP (driving gRNA) elements target both chromosome homologs in almost exactly same location, a situation most often favoring the receiver allele, which, unlike the Cas9 transgene, is flanked by perfect homology on both sides. Finally, ERACRs with two gRNAs targeting neighboring sites on the same chromosome homolog generated an abundant class of damaged chromosomes. This effect was not the result of dual cutting per-se, since single-cut ERACRs generated comparable frequencies of damage. Thus, one-sided homology mismatch rather than dual cutting of the target most likely underlies ERACR-induced chromosome damage.

e-CHACRs and ERACRs perform effectively as designed in cage experiments

Despite imperfections revealed in pair-matings, e-CHACRs and ERACRs performed largely as intended in multi-generational cage studies. In these experiments, e-CHACR-wR completely eliminated Cas9 activity and drove itself to 100% homozygous introgression in 9–10 generations. Similarly, ERACR-2 entirely replaced the MCR-GFP element over 6 generations. Several factors contributed to these successful outcomes (Fig. 8B): 1) efficient Cas9 mutagenesis by the targeting gRNA, 2) efficient e-CHACR-wR copying, and 3) absence of somatic mosaicism or shadow-drive caused by accumulated Cas9/gRNA complexes (this factor, which greatly reduces formation of drive-resistant NHEJ alleles, likewise pertains to ERACR performance). Also, a fitness cost may be associated with the MCR-GFP element, inferred from modeling and competition between the MCR-GFP element and a y⁻ allele in cages (Fig. S22A). In addition, single generation crosses revealed that e-CHACRs can damage MCR-GFP chromosomes, generating a class of GFP⁻ male-lethal alleles. These alleles would be invisible in the cage experiments since they do not survive either in males or homozygous females but would deplete the MCR-GFP allelic pool.

Three notable features distinguished the drive trajectories of the y⁻ ERACR-min and y⁺ ERACR-2, all of which derive from their y⁻ versus y⁺ phenotypes. First, ERACR-2 rapidly drove to 100% replacement, while ERACR-min plateaued and then increased more slowly without entirely eliminating the MCR-GFP element. Second, inter-cage variation was much reduced for the ERACR-2 trials than for ERACR-min, where the observed variation was more typical (~5–10%). Finally, ERACR-2 experienced a one generation lag in producing progeny carrying both the DsRed+ ERACR and MCR-GFP elements (yellow curves in Fig. 7D versus Fig. 7F).

The significant assortative mating advantage conferred upon y⁺ ERACR-2 males relative to y⁻ MCR-GFP males is well-documented (Barker, 1962; Bastock, 1956; Diederich, 1941; Merrell, 1949; Sturtevant, 1915), and was confirmed in our own experiments (Fig. S23A–E). Indeed, the y⁺ ERACR-2 allele overtook a point-mutant y⁻ allele in competitive cage experiments (Fig. S23E), albeit more gradually than when driving in a Cas9-dependent fashion against the MCR-GFP element (Fig. 7F). The more rapid and complete drive of ERACR-2 compared to ERACR-min can be attributed to dual (Cas9-mediated + assortative mating) versus single (Cas9-mediated) drive mechanisms being operative. Similarly, the initial delay in generating ERACR-2/MCR-GFP heterozygotes is expected if y⁺ ERACR-2 females preferentially mated with their own kind. The much-reduced inter-cage variation observed for the ERACR-2 trajectories may reflect two independently acting processes: drive produced by Cas9 delivered in-trans and a selective pressure resulting from assortative mating favoring y⁺ males. These dual factors provide fewer opportunities for stochastic variation to accumulate.

Strengths of ERACR and e-CHACR neutralizing elements:

These studies provide encouraging support for neutralizing a gene-drive with e-CHACRs or ERACRs. The major strength of e-CHACRs is that they act generically and can neutralize SpCas9 gene-drives inserted at various locations in the genome. e-CHACRs can efficiently inactivate Cas9 and copy themselves (e-CHACR-wG/R elements). The primary advantage of ERACRs is that they can delete and replace the gene-drive, thereby eliminating any undesired effect associated with that element. ERACRs perform as intended less often, replacing the gene-drive as designed only about a third of the time (Fig. 4A–B). Nonetheless, ERACRs, particularly ERACR-2 carrying recoded sequences to restore target locus function, performed very effectively in population cages.

In the future, these two neutralizing strategies could be combined to exploit the strengths of each system. These systems also could be implemented with other proposed mitigation measures such as elements carrying anti-Cas9 proteins or inundative releases of strains carrying functional cleavage-resistant alleles of the targeted locus. Such passive elements could spread rapidly if insertion of a gene-drive disrupted locus function. Also a CATCHA construct has been described that is inserted at the site of the gene-drive (similar to ERACRs), and is designed to copy into the Cas9 transgene (Wu et al., 2016). While CATCHA elements could target Cas9 sources located elsewhere in the genome, they would not copy in those contexts. Also, because the cassette must be flanked with Cas9 homology arms, it does not readily deliver an in-frame recoded target gene.

Another potential strategy is to incorporate a “weakened” gRNA* targeting the same essential Cas9 amino acid residues as gRNA-C1 or gRNA-C3 into the drive element itself. gRNA*s carrying one or two base-pair mismatches to the target sequence should reduce Cas9 cleavage by 1–2 logs. While not expected to appreciably interfere with the spread of a gene-drive carrying such a gRNA*, the Cas9 transgene should be mutated at rates ~10⁵ greater than by spontaneous mutation once the drive attains full introgression into a population, thereby accelerating elimination of Cas9 from the population if it imposed a fitness cost.

Implications for the potential implementations of neutralizing systems:

While e-CHACR and ERACR elements behave largely as expected to curtail the spread of gene-drives, they also produce unexpected outcomes (e.g., generation of drive-competent chimeric ERACR/MCR-GFP elements) and can falter in copying (e.g., e-CHACRs with non-optimal gRNAs). Thus, while these experiments provide optimism for strategies to retard wayward gene-drive systems if necessary, we agree with the recommendations of the National Academy of Sciences Report on Gene-Drives (Organisms, 2016) regarding the cautious use of such neutralization systems. We concur with their appraisal that the decision to go forward with potential releases of gene-drive systems should not be predicated on constructing neutralizing elements, and that such systems should be developed only for precautionary purposes. If one has concerns about a potential gene-drive system somehow going awry, then surely such concerns would only be amplified by release of a second element which could generate yet more complex genetic outcomes. Nonetheless, the current in-depth studies provide encouraging support for the use of these types of mitigating strategies should there be an unanticipated need for them.

STAR Methods

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ethan Bier (ebier@ucsd.edu).

Materials Availability

All unique plasmids or Drosophila lines generated by this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Data and Code Availability

Raw counting data from Drosophila melanogaster fly crosses, construct maps, and all sequencing data/chromatograms have been deposited on Mendeley Data (doi: 10.17632/hjcpd6j8rn.1).

Experimental Model and Subject Details

Drosophila rearing and Genetic Experiments

All genetic experiments were conducted in a high-security Arthropod Containment Level 2 (ACL2) barrier facility, in accordance with protocols approved by the Institutional Biosafety Committee from University of California San Diego. All materials and waste are frozen for 48 hours prior to removal from the facility followed by autoclaving. Flies were kept triple-contained in shatter-proof polypropylene plastic vials. Stocks were maintained at 18°C while crosses were carried out at 25°C on a 12 hour day/night cycle on cornmeal media.

Multi-Generational Population cage studies

All population cage experiments were conducted at 25°C with a 12 hour day-night cycle, in 250 ml bottles containing standard cornmeal medium. Seeding populations for all drive experiments included equal numbers of unmated females and males for each genotype. 5–7 days following introduction into the cage and successful mating, all flies (generation n) are removed. Subsequent progeny (generation n+1) were collected, then randomly separated into two pools, one group is further analyzed and screened while the other pool is used for further seeding of the next cage (generation n+2).

Method Details

Plasmid Construction

All Constructs were cloned using standard recombinant DNA techniques, including PCR amplification with Q5 Hotstart master mix (NEB #M0494S) and Gibson assembly with NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs, Cat. # E2621) . Following successful cloning, the plasmids were transformed into NEB 5-alpha Electrocompetent Competent E. coli (New England Biolabs, Cat. # C2989). We list starting plasmids, oligos, and restriction enzymes used to generate each construct. Refer to supplementary materials (Table S2) for full sequences of plasmids and oligos.

MCR-GFP construct by “tagging”

The MCR-GPF construct used for these studies was generated by genetic “tagging” of the unmarked prototype yMCR construct (Gantz and Bier, 2015). Briefly, a plasmid construct carrying the eGFP gene expressed under control of the 3XP3 eye promoter, followed by 3’ SV40 polyadenylation sequences, flanked by homology sequences present in the unmarked yMCR construct (Cas9 transgene and U6 gRNA homology arms), and carrying a construct expressing a gRNA directing cleavage between the Cas9 transgene and U6-gRNA of the yMCR was inserted into the genome at the same site as the yMCR using gRNA-y1. Transgenic fly strains carrying this tagging construct (eGFP-Tagger) were recovered and crossed to the yMCR line to generate MCR/eGFP-Tagger females. Single crosses were performed by mating with Basc balancer males and the daughters in the resulting offspring were screened for a y- body color, indicative of Cas9 activity. Vials with Cas9 activity indicated a successful tagging of the yMCR construct and individual males were crossed with cut resistant gRNA-y1 site females (contain static Cas9 at gRNAy1 site) to generate isogenic homozygous stocks for use in this study.

Microinjection of ERACR and e-CHACR Constructs

Plasmids were purified using the Qiagen Plasmid Midi kit (#12191). ERACR and e-CHACR constructs were co-injected with a transient source of pAct-Cas9 (pAct-Cas9 was a gift from Fillip Port (Addgene plasmid # 62209) by Best Gene Inc. Injection mixes were assembled with either ERACR or e-CHACR donor plasmids (final concentration: 700 ng/μl) and pAct-Cas9 (final concentration: 300 ng/μl) in a volume of 50 μl. ERACR mixes were injected into a w1118 stock (BDSC #5905) while e-CHACR mixes were injected into an Oregon-R stock (BDSC #2376). All constructs were fully sequenced prior to injection and subsequently generated transgenic stocks were confirmed through sanger sequencing as well.

Recovery of Transformants and Genomic DNA Isolation

Injected flies were crossed the w1118 stock and progeny were screened for DsRed or GFP positive individuals, which were then subsequently crossed to an FM7 (first chromosome) or TM3 (third chromosome) balancer. Homozygous lines were established based on the absence of balancer alleles. For analysis of individuals containing gene-drive elements, flies were frozen in an ACL-2 facility for 48 hours prior to removal from the facility and DNA isolation. Genomic DNA was prepared from individual male flies according to protocols by (Gloor et al., 1993).

Drosophila PCR and Sequence Analysis

Sequencing PCR reactions were assembled with Q5 Hotstart master mix (NEB #M0494S), KOD Xtreme Hot Start DNA Polymerase (Millipore #71975), Phusion High-Fidelity PCR Master Mix with HF Buffer (Thermo-Fisher #F531S), or Platinum SuperFi DNA Polymerase (ThermoFisher #12351010). Primers used to amplify PCR products are listed in Table S4. PCR samples were purified using the QIAquick PCR purification kit (Qiagen #28104) prior to sanger sequencing.

Quantification and Statistical Analysis

Mathematical Modeling

Model fitting was carried out for the ERACR-min and ERACR-2 cage experiments using Markov chain Monte Carlo (MCMC) methods in which genotype-specific fitness parameters were estimated, including 95% credible intervals (CrIs). We considered discrete generations and Mendelian inheritance rules at the gene drive locus, with the exception that, for females heterozygous for the ERACR (E) and homing (H) allele, a proportion of the homing alleles are cleaved, while a proportion, , remain H alleles (Fig. Smod1). Of those that are cleaved, a proportion are subject to accurate homology-directed repair (HDR) and become E alleles, while a proportion become resistant alleles. Of those that become resistant alleles, a proportion become in-frame, cost-free resistant (R) alleles, while the remainder become out-of-frame or otherwise costly resistant (broken, B) alleles. Parameter values were fixed for and based on the experimental results depicted in Fig. S12. These considerations allowed us to calculate the expected genotype frequencies in the next generation, and to explore the fitness and assortative mating parameters that maximize the likelihood of the experimental data. Estimated parameters include the fitness cost associated with males having the homing allele females homozygous for the homing allele and having one copy of the homing or broken allele in females or respectively. BB females and BY males were known to be unviable. Populations were assumed to be randomly mixing, with the exception of the ERACR-2 experiment, in which females having the y+ phenotype preferentially mated with males having the y+ phenotype. For ERACR-2, we estimated an additional parameter representing the fraction by which y+ females reduce their mating with y- males. The modeling framework is described in full in the Supporting Information.

Statistical Methods

Statistical analysis was performed using GraphPad Prism. One-way independent measures ANOVA followed by Sidak’s Multiple Comparisons test were performed for graphs in Figure 1D and Figure 6 (all panels). All other graphs use one-way independent measures ANOVA followed by Tukey’s post-hoc test. Significance was determined as follows: * denotes p<0.05, ** denotes p<0.01, *** denotes p<0.001, and **** denotes p<0.0001.

Supplementary Material

2

Table S4: PCR primers for sequence analysis for Figures S6 and S14, related to STAR methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
NEB 5-alpha Electrocompetent Competent E. coli	NEB	C2989
Chemicals, Peptides, and Recombinant Proteins
Q5 Hotstart High-Fidelity 2X Master Mix	NEB	M0494S
Phusion High-Fidelity PCR Master Mix with HF Buffer	Thermo-Fisher	F531S
Platinum SuperFi DNA Polymerase	Thermo-Fisher	12351010
KOD Xtreme Hot Start DNA Polymerase	Millipore	71975
Critical Commercial Assays
Qiagen Plasmid Midi Kit	Qiagen	12191
QIAquick PCR purification kit	Qiagen	28104
NEBuilder HiFi DNA Assembly Master Mix	NEB	E2621
Deposited Data
Raw data (counts from fly experiments, construct maps, sequencing chromatograms)	Generated by this study	doi: 10.17632/hjcpd6j8rn.1
Experimental Models: Organisms/Strains
Oregon-R	Bloomington Drosophila Stock Center	BDSC #5905
W1118	Bloomington Drosophila Stock Center	BDSC #2376
FM7a	Bloomington Drosophila Stock Center	BDSC #785
Basc	Bloomington Drosophila Stock Center	BDSC #806
w[a]	Bloomington Drosophila Stock Center	BDSC #148
ac[4] w[a]	Bloomington Drosophila Stock Center	BDSC #8
w[*]; TM3, Sb[1] Ser[1]/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC#2537
y-; w- (used in MCR cage trial)	Gift from the Gantz Lab	N/A
y-; w- (F5 line used in ERACR cage trial)	Gantz et al., 2015	N/A
Vasa Cas9 in yellow (y1) w[1118A]	Gift from the Gantz Lab	N/A
Vasa Cas9 in yellow (y1) w+	Gift from the Gantz Lab	N/A
Vasa Cas9 in white (w2)	Gift from the Gantz Lab	N/A
w- MCR GFP	This study	N/A
w[a] MCR GFP	This study	N/A
w+ MCR GFP	This study	N/A
e-CHACR-AG; see Table S3	This study	N/A
e-CHACR-AR; see Table S3	This study	N/A
eCHACR-e; see Table S3	This study	N/A
e-CHACR-k; see Table S3	This study	N/A
vCas9-R; see Table S3	This study	N/A
ERACR-min; see Table S3	This study	N/A
ERACR-1; see Table S3	This study	N/A
ERACR-2; see Table S3	This study	N/A
ERACR-2 y2 Single-cut; see Table S3	This study	N/A
ERACR-2 y3 Single-cut; see Table S3	This study	N/A
X-duplication lines; see Table S4	Bloomington Drosophila Stock Center	Various; See Table S5
Oligonucleotides
Cloning/construct design primers; see Table S2	N/A	N/A
PCR primers; see Table S4	N/A	N/A
Recombinant DNA
ERACR-min	This study	N/A
ERACR-1	This study	N/A
ERACR-2	This study	N/A
ERACR-2 y2 single cut	This study	N/A
ERACR-2 y3 single cut	This study	N/A
e-CHACR-AG	This study	N/A
e-CHACR-AR	This study	N/A
e-CHACR-e	This study	N/A
e-CHACR-k	This study	N/A
Vasa Cas9 resistant to e-CHACR-e	This study	N/A
pAct-Cas9	Gift from Fillip Port	Addgene # 66209
Software and Algorithms
GraphPad Prism	GraphPad	https://www.graphpad.com/scientificsoftware/prism/
Snapgene	Snapgene	https://www.snapgene.com
Microsoft Excel	Microsoft	N/A

Highlights.

1. e-CHACRs can efficiently copy and inactivate Cas9 activity (~99%).

2. e-CHACRs spread to 100% prevalence in cage trials and eliminate Cas9 activity.

3. ERACRs often copy, but can also damage the target chromosome.

4. ERACRs can efficiently delete and replace a gene-drive in population cages